Near-field laser-induced fluorescence velocimetry of
neutral xenon in a 6-kW Hall thruster plume
IEPC-2007-229
Presented at the 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
Timothy B. Smith∗, Wensheng Huang†, Bryan M. Reid ‡,
and
Alec D. Gallimore§
Plasmadynamics and Electric Propulsion Laboratory,
University of Michigan, Ann Arbor, MI 48105, USA
We present laser-induced fluorescence (LIF) measurements of neutral xenon (6s 2 [3/2]02 )
in the near-field plume of a 6-kW Hall thruster. The 823.163 nm (air) beam from a TUI
Optics TA-100/830 tapered-amplifier diode laser is focused along an axial beamline onto
a submillimeter interrogation volume, from which we collect resonant LIF. Polarization
control via a rotating half-wave plate keeps the input beam polarization parallel to the
magnetic field, minimizing distortion of the LIF spectrum by Zeeman splitting in regions
of high magnetic field intensity. Doppler velocimetry of LIF spectra taken at the discharge
channel exit returns bulk axial velocities at discharge voltages of 53 to 300 V and mass
flow rates of 5 through 17.5 mg/s.
I.
Introduction
Considerable effort is being invested into understanding the role of neutral flow dynamics in the operation
of conventional Hall thrusters. In particular, understanding the near-field neutral velocity will provide insight
into the effects of anode injection schemes, wall accommodation, and resonant charge-exchange collisions in
the near-field plume. A comprehensive review of past neutral flow dynamics research indicates that anode
injection schemes can significantly affect thruster efficiency, lifetime, stability, and thermal margin.1 In
response to these potential benefits, the neutral flow dynamics in the discharge channel are being altered to
maximize propellant utilization by increasing the neutral flow uniformity and residence time. By optimizing
the neutral flow dynamics in the discharge channel, the operational envelope of Hall thrusters can be expanded
to high specific impulse and high thrust-to- power. At these challenging operational regimes, the importance
of propellant utilization and thruster stability are amplified.
To track the effects of changes in neutral flow dynamics, neutral properties will be tracked in the verynear-field (< 1 thruster diameter). Measuring the neutral component of partially ionized plasmas is relatively
difficult with electrostatic probes; however, non-intrusive optical methods are well suited for this application.
This experiment uses Laser-Induced Fluorescence (LIF) velocimetry of neutral xenon on a planar interrogation region extending downstream of the discharge channel exit. Correlation of the velocity profile with
analytical models, plume simulations, and electrostatic probe measure will permit a thorough investigation
of the effects of neutral flow non-uniformity and residence time.
1
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
129Xe
I = 1/2
131Xe
I = 3/2
5/2
F' = 3/2
J' = 2
5/2
F = 3/2
J=2
7/2
5/2
3/2
1/2 = F'
7/2
.
.
(a) Fine structure.
5/2
3/2
1/2 = F
(b) Hyperfine structure for absorption line.
Figure 1. Optogalvanic and LIF schemes using Xe I absorption at 823. nm.
II.
Theory
Figure 1(a) shows the fine structure transitions for optogalvanic and resonant LIF spectroscopy of the
823.163 nm (air) transition of Xe I. Absorption of 823.163 nm laser light causes xenon atoms resting in
the metastable lower state 6s 2 [3/2]02 to jump to the excited upper state 6p 2 [3/2]2 . Since this upper state
is closer to the first ionization potential for xenon (12.1298 eV) than the lower state, the probability of
electron-impact ionization (and thus the galvatron discharge current) increases linearly with the upper state
population, providing an optogalvanic signal. Those atoms which are not ionized out of the upper state
quickly decay; since there are no other allowed transitions from the 6p 2 [3/2]2 , the 6p 2 [3/2]2 → 6s 2 [3/2]02
fluorescence at 823.163 nm provides a strong, but resonant LIF signal.
A.
Doppler shift and broadening
Consider a light source with vacuum wavelength λ and frequency ν = c/λ. An observer who is stationary
with respect to the light source will see light at the same frequency ν. An observer moving towards the light
source will see a bluer (i.e., higher-frequency) light than the stationary viewer, while an observer moving
away from the light source will see a redder (i.e., lower-frequency) light.
In LIF, this Doppler effect appears as a shift in the resonant frequency ν0 as the laser is scanned over
a very short frequency range. The change in photon frequency 4ν = ν − ν0 for a particle with velocity v
passing through a light beam of wave vector k is
4ν = −
k·v
.
2π
(1)
Given a beamwise velocity component vk ≡ v · k̂, Eqn. 1 gives the beamwise velocity-to-frequency transformations
vk ν =
1−
ν0
(2)
c ν
vk =
1−
c.
(3)
ν0
A swarm of particles with a normalized velocity distribution f (v) will also “see” the frequency of incoming
photons shifted by the relative velocity of the particle in the direction of the photon. The resulting Doppler
∗ Research
Fellow and Lecturer, Aerospace Engineering, timsmith@umich.edu
Research Assistant, Aerospace Engineering, davhuang@umich.edu
‡ Graduate Research Assistant, Aerospace Engineering, reidb@umich.edu
§ Professor and Associate Dean, Aerospace Engineering, rasta@umich.edu
† Graduate
2
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
lineshape will be shifted by the beamwise bulk velocity uk = hv · k̂i and broadened by the thermal width of
the beamwise distribution f (vk ). The generalized Doppler lineshape d(ν), when properly normalized so that
Z ∞
d(ν) dν = 1,
(4)
−∞
is given by
c
d(ν) = f
ν0
ν
1−
c .
ν0
(5)
In species with no hyperfine structure (HFS), the LIF spectrum i(ν) is the convolution of the Doppler
broadening with the natural (or lifetime) broadening function
l(ν) =
4νn
1
2π (ν − ν0 )2 + (4νn /2)2
(6)
where 4νn = Aj /(2π) is the natural linewidth for state j. This Lorentzian function is usually much narrower
for warm gases than the Doppler broadening. In this case, the natural broadening can be replaced by a Dirac
delta function,
i(ν) = d(ν) ⊗ l(ν) ≈ d(ν) ⊗ δ(ν) = d(ν),
(7)
letting us directly transform the LIF spectrum into an excellent approximation of the beamwise velocity
distribution.
B.
Hyperfine structure
In xenon (and other species with HFS), the LIF spectrum can be modeled as the convolution of the hyperfine
splitting function h(ν) with natural and Doppler broadening functions,
i(ν) = h(ν) ⊗ l(ν) ⊗ d(ν).
(8)
The hyperfine splitting h(ν) is a series of n Dirac delta functions,
h(ν) =
n
X
pj δ(ν − νj ).
(9)
j=1
where νj is the j th hyperfine line center and pj is the j th hyperfine line intensity. The 6s 2 [1/2]01 → 6 2 [3/2]2
transition of Xe I has n = 18 such hyperfine splittings.
Hyperfine structure consists of two components. Isotopic splitting is caused by the mass and nuclear
volume differences between isotopes in the sample. Nuclear-spin splitting is caused by the coupling between
the nuclear spin I and the total electronic angular momentum J. For the sake of brevity, we will henceforth
use the term “HFS constants” to denote the combined set of isotopic shifts and nuclear-spin interaction
constants.
There are nine stable isotopes of xenon, seven of which have natural abundances greater than one percent.
Each of these isotopes causes a slightly different transition energy Eij , resulting in isotopic splitting. Table
1 presents naturally-occurring xenon isotopic abundances and shifts (conventionally defined relative to the
center frequency for 132 Xe).
Table 1. Isotopic abundance and shift relative to 132 Xe for the Xe I 6s 2 [3/2]02 → 6p 2 [3/2]2 transition at 823.163
nm (air) nm (from Jackson and Coulombe et al.4 ). No isotopic shift values are available for 124 Xe and 126 Xe.
Mass (amu)
Abundance (%)
Shift (MHz)
124
0.0096
—
126
0.009
—
128
1.92
84.1
129
26.4
96.5
130
4.1
37.1
131
21.1
72.8
132
26.9
0.0
134
10.4
-38.9
136
8.9
-116.9
Seven of the nine stable isotopes of xenon have even atomic mass, resulting in no nuclear spin I. The two
isotopes with an odd atomic mass, however, have non-zero nuclear spin quantum numbers I. The lighter
3
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
isotope, 129 Xe, has I = 1/2, while 131 Xe has I = 3/2. These non-zero nuclear spins cause nuclear-spin
splitting of the atomic energy levels. This nuclear-spin splitting is 1 to 2 orders of magnitude broader than
the isotopic splitting, and provides most of the characteristic shape of our absorption line.
The total angular momentum F = I + J takes quantum number values
F = I + J, I + J − 1, . . . , |I − J|
(10)
where J is the total electronic angular momentum quantum number.9 The extra term energy due to nuclearspin splitting is given by8
C
(11)
Enss = A + BD,
2
where A is the nuclear magnetic dipole interaction constant, B is the nuclear electric quadrupole interaction
constant, and the terms
C = F (F + 1) − I(I + 1) − J(J + 1)
(12)
and
D=
(3C/4)(C + 1) − I(I + 1)J(J + 1)
2I(2I − 1)J(2J − 1)
(13)
contain the nuclear spin-orbit interactions.
Table 2 presents energy levels and nuclear-spin structure constants for the Xe I absorption line at 823.163
nm (air), taken from Jackson and Coulombe.6
Table 2. Energy levels and nuclear-spin structure constants for the Xe I 6s
823.163 nm.
Level
6s 2 [3/2]02
6p 2 [3/2]2
Ei
(eV)
8.353155
9.8210934
A129
(MHz)
−2385
−892
A131
(MHz)
706
264
2 [1/2]0
2
→ 6
2 [3/2]
2
transition at
B131
(MHz)
253
30
The transition rule for nuclear-spin splitting is 4F ≡ F − F 0 = [0, ±1], where F is the upper and F 0
is the lower state’s total angular momentum quantum number. (The zero-zero transition is forbidden, F =
0 6→ F 0 = 0.) Figure 1 shows the fine structure and nuclear-spin splitting for he Xe I 6s 2 [1/2]02 → 6 2 [3/2]2
transition at 823.163 nm.
The relative intensity of each nuclear-spin split component is given for a J → J − 1 transition by10
I(F → F − 1)
∝
I(F → F ) ∝
I(F − 1 → F ) ∝
P (F )P (F − 1)
F
(2F + 1)
P (F )Q(F )
F (F + 1)
Q(F )Q(F − 1)
F
(14)
(15)
(16)
where P (F ) = (F + J)(F + J + 1) − I(I + 1) and Q(F ) = I(I + 1) − (F − J)(F − J + 1).
For the isotopes with even mass numbers (i.e., without nuclear-spin splitting), the line intensity pj is
linearly proportional to the naturally-occurring abundance for each isotope. The line intensities of 129 Xe
and 131 Xe are linearly proportional to the product of the isotopic abundance and the relative intensity of
the nuclear-spin split components.
C.
Propellant utilization and ionization fraction
Neglecting facility effects, the propellant utilization determines the ionization fraction and neutral number
density in the near-field plume of a Hall thruster. Since the Xe I LIF signal strength and signal-to-noise ratio
are roughly proportional to the neutral number density, propellant utilization has an undeniable influence
on the effectiveness of Xe LIF in the Hall thruster plume.
4
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
In most analyses, the propellant utilization is defined as the ratio of ion flux to the total particle flux.
For an atomic species with Z electrons, a generalized expression for propellant utilization is
, Z
Z
X
X
ηu =
ni ui
ni ui
(17)
i=1
i=0
where ni is the number density and ui is the axial velocity for species i (and we define i = 0 for neutrals,
i = 1 for single ions, i = 2 for double ions, and so forth). Given the total number density
n = n0 + n1 + . . . =
Z
X
ni
(18)
i=0
and defining the ionization fraction
αi = ni /n
(19)
for each ionization state 1 ≤ i ≤ Z, the generalized propellant utilization can be expressed in terms of
ionization fraction as
, Z
Z
X
X
ηu =
αi ui
αi ui .
(20)
i=1
i=0
When multiple ionization is negligible, the propellant utilization becomes
(ηu )1 =
n1 u1
n0 u0 + n1 u1
(21)
or, in terms of the ionization fraction α1 = n1 /n,
1 − α1
(ηu )1 =
α1
u0
u1
−1
+1
.
(22)
Equation 22 can be rearranged to yield the ionization fraction as a function of propellant utilization, neutral
velocity and ion velocity:
−1
1 − (ηu )1 u1
+1
.
(23)
α1 =
(ηu )1
u0
Since α0 + α1 = 1 for negligible multiple ionization, Equation 23 shows that the neutral number density n0
(and thus, the LIF signal strength and SNR) varies as
−1 !
1 − (ηu )1 u1
n0 = n 1 −
+1
(24)
(ηu )1
u0
Thus, the LIF signal intensity (ceteris paribus)
• decreases with increasing ion velocity u1 , and
• decreases with increasing propellant utilization (ηu )1 .
In the limit of 100% propellant utilization, the ionization fraction goes to 100% and the Xe I LIF signal (and
SNR) drop below the threshold of detection.
III.
A.
Apparatus and Procedure
Facility
Experiments were performed in the Large Vacuum Test Facility (LVTF) at the University of Michigan. The
LVTF is a φ6 m ×9 stainless steel-clad cylindrical chamber. Pumping is provided by seven single-stage
CVI model TM-1200 cryopumps and liquid nitrogen shrouds, with a nominal pumping speed of 240,000 l/s.
Facility pressure is monitored by two hot-cathode ionization gauges, and base pressure is approximately 1.2×
5
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
10−7 Torr. The facility pressure during thruster operation at 11 mg/s total xenon flow rate is approximately
7.6 × 10−6 Torr, corrected for xenon.
Xe I LIF spectra were recorded at the discharge channel and cathode exit planes of a 6-kW laboratory
model Hall thruster. During these experiments, the 6-kW Hall thruster was mounted on two crossed translation tables. The axial translation table has a 1-m range of motion, while the radial translation table has
a 2-m travel. Both translation stages are PC-controlled by a custom LabView VI, with locational resolution
on the order of 0.1 mm. A vertically-adjustable support structure atop the translation tables elevated the
thruster to the chamber centerline, allowing the plume to expand ∼ 4 meters along the chamber axis.
Research-grade xenon propellant (99.999% pure) was supplied to the thruster by separate commercially
available flow meters and controllers, having an accuracy of ±1% of full scale. Calibration of the flow system
by the constant volume method2 includes the effects of compressibility.3
B.
Laser and optics
The laser system used in this experiment is a TUI Optics TA-100/830 tapered-amplifier diode laser. This
system has a nominal linewidth of ∼ 10 MHz, a tuning range of 820 nm to 836 nm, and a mode-hop-free
range of over 10 GHz. Though capable of output powers in excess of 500 mW, off-peak efficiency drops and
Faraday isolation of the tapered amplifier reduce the available output power at the upper end of the system
wavelength range (835 ± 1 nm) to approximately 250 mW.
The master oscillator for this system is an ECDL. Coarse wavelength control is provided by Peltier cooling
of the ECDL and micrometer adjustment of the ECDL feedback grating angle. PC control of the fine tuning
is provided by a LabVIEW VI, which uses the auxiliary output of a Stanford SR810 lock-in amplifier to send
a control voltage to a piezoelectric transducer (PZT) driving the ECDL feedback grating. Feedback circuitry
in the ECDL power supply decreases the master oscillator diode current during a PZT sweep to maximize
the mode-hop-free range. In order to include the effect of changing laser output power, we normalize our
reported LIF and OGS spectra by a quadratic curve-fit to the master oscillator power. It is important to note
that this normalization carries an implicit assumption that the absorption line is not saturated. Steering
mirrors inside the TA-100 enclosure couple the ECDL output into the tapered amplifier diode, which boosts
the ECDL output power by a factor of more than 20 while maintaining the master oscillator wavelength and
linewidth.
Although power changes over the course of a frequency scan, simply normalizing the signal by the laser
power would give misleading results when the LIF lineshape is highly saturated. Instead, we use an external
photodiode to monitor laser power, and take care to keep the central peak of the photodiode signal within
a 5% tolerance band for all scans. This allows a basis for comparing the magnitude of the signal between
multiple scans at different operating conditions.
Figure 2 shows a schematic of the diode laser table contents. A wedged beam pickoff diverts two beamlets
(beams A and B, each ∼ 5 mW) from the main TA-100 output beam. The main beam passes through an
initial beam splitter, which diverts ∼ 40% of the main beam to a second beam splitter that further dividing
the beam into two ∼ 50 mW beamlets (beams C and D). The remaining ∼ 120 mW of main beam power
(beam E) continues on toward the LVTF.
1.
2.
3.
4.
5.
6.
7.
8.
9.
Index:
Tunable diode laser
2-GHz etalon assy
Wavemeter
Half-wave plate
Polarizing beam cube
Chopper
Neutral density filter
Photodiode
Optogalvanic cell
Figure 2. Optical table schematic for 6-kW Hall thruster LIF experiment.
6
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
Beam A goes through a Stanford SR540 chopper running at 1.6 kHz to two mirrors, which steer the
beam through a neutral density filter, into a Hamamatsu L2783 galvatron, and terminates at a Thorlab
DET110 photodiode. The optogalvanic current and photodiode voltage signals go to Stanford SR810 lockin amplifiers, yielding optogalvanic and absorption spectra for the galvatron discharge. During saturation
studies, a half-wave plate, polarizing beam splitter cube, and a neutral density filter are inserted in the beam
path to control the laser power entering the galvatron.
Beam B goes through same SR540 chopper, and passes through a neutral density filter to an external
photodiode. Since the photodiode has a known efficiency and is connected to a preamplifier of known
characteristics, we can make highly precise measurements of the beam power.
Beam C is steered by two mirrors through a variable neutral density filter to the Burleigh WA-1000
wavemeter. The wavemeter can determine laser wavelength with a 1 pm resolution, which is sufficient for
coarse alignment and troubleshooting of the laser. Oversampling of the PZT control voltage permits greater
wavelength resolution by a quadratic curve-fit to the wavemeter output, but does not improve the absolute
bias of the laser wavelength measurement.
Beam D is directed by two mirrors through an iris into a Burleigh SA-91 etalon assembly with a free
spectral range of 2 GHz and a finess of 256. This provides a series of sharp frequency markers, allowing
high-resolution (∼ 10 Mz) measurement of the frequency interval swept during a laser scan. Sampling steps
were never set this small during this experiment, though; typical sampling intervals ranged between 25 and
50 MHz. Cost and time limitations were the major driver for this reduced frequency resolution, but the need
to reduce thermal drift of the etalon spacing was also factor.
Beam E continues on through a half-wave plate, a polarizing cube, and a Stanford SR540 optical chopper
running at 1700 kHz before reaching the first of 3 near-infrared mirrors. These mirrors steer the beam
through the upper sapphire window, into the LVTF, and onto the input window of the LVTF periscope.
Laser input window
Spex H-10
Periscope
Beam from diode laser
Collection lens
Focused laser beam
Collimated fluorescence
Collection lens
(a) Looking upstream (north).
Laser focus lens
Translation table
(b) Looking west.
Figure 3. Laser beam delivery and fluorescence collection optics in the LVTF.
Figures 3(a) and 3(b) show the LVTF beam handling setup. Approximately 50 mW of beam power enters
the periscope through a replacable antireflection (AR) coated window. Three near-infrared mirrors within
the periscope direct the beam to a focusing telescope parallel to the cathode and solenoid axis, reducing the
beam diameter to less than 1 mm.
Two separate AR windows protect the φ100 mm, f /2.5 collection lens. The collimated fluorescence from
the thruster plume passes out of the LVTF through another sapphire window and into a φ100 mm, f /5 lens.
Two protected aluminum mirrors direct the focusing beam onto a Spex H-10 monochromator (using 500-µm
slits) with a Hamamatsu 928 PMT. A Stanford SR850 lock-in amplifier extracts the LIF signal from the PMT
output. The interrogation volume, defined by the intersection of the laser beam and the doubly-magnified
monochromator slit image, is a φ1 × 1 mm cylinder oriented parallel to the cathode axis.
C.
Optogalvanic cell
The optogalvanic cell used in this experiment is a Hamamatsu L2783-42 XeNe-Mo galvatron. The core of
the galvatron is a φ6.25 mm pair of cylindrical Mo electrodes, centered in a φ25 x 120 mm glass cylinder
7
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
filled with approximately 3 Torr of xenon and and 4 Torr of neon. The ends of the cylinder are angled
approximately 10 degrees from a perpendicular to the electrode axis; though this is less than Brewsters
angle for a glass-air interface, it reduces the likelihood of retroreflected beams entering the discharge. This
design permits unobstructed passage of a laser beam along the axis shared by the cylinder and electrodes.
A 192 V discharge across the galvatron electrodes produces a reasonably cool, dense, stationary plasma.
Additionally, this voltage seems to maximize the population of the 6s 2 [3/2]02 metastable and provide clean
optogalvanic signal. A ballast resistor prevents runaway current growth after the discharge is struck, and
provides a simple method of measuring the discharge current. The voltage drop across the ballast resistor is
connected to a Stanford SR-810 lock-in amplifier through a 3.33 µF, 450 V capacitor. The capacitor passes
the AC component of the voltage drop, while isolating the lock-in from high DC voltages.
IV.
Results
Our original plan was to take Xe I LIF spectra at the cathode exit and in a region downstream of discharge
channel. After taking scans at multiple radial locations across the discharge channel exit, we planned to
continue downstream until the metastable lower-state density became too low for good LIF signal-to-noise
ratios (SNRs).
Unfortunately, we quickly found that at normal operating conditions for this Hall thruster (discharge
voltage Vd ≥ 300V and anode mass flow ṁA ≥ 10 mg/s), the neutral density was so low that acceptable
signal-to-noise ratios (SNR> 1) required excessively-long scans (T ≥ 3600 s). Since Equation 23 implies that
the neutral density will tend to increase with decreasing discharge voltage, we decided to carry a survey of
discharge voltages below 300 V. In the course of investigating these low-voltage discharges, we discovered
that anode mass flow rate also affects the LIF signal quality.
A.
Zeeman splitting
(a) B = 0 gauss
(b) B = 174 gauss
Figure 4. Optogalvanic (black line) and cathode exit plane LIF (red line) spectra at 823.163 nm (air), cathode
operating without an anode discharge.
Figures 4(a) and 4(b) show how an axial magnetic field at the cathode exit causes Zeeman splitting of
the LIF spectrum at 823.136 nm. The SNR is quite good for both spectra, largely because the discharge
current is off. Because the magnetic field has a strong component parallel to the beam vector (B k k) and
the beam vector is always perpendicular to the polarization vector (mbE ⊥ k), it is impossible in this case
to remove Zeeman splitting by rotating the polarization vector into π-polarization.
Figures 5(a) and 5(b) show how increased discharge current adds noise to the cathode LIF spectrum.
The Zeeman splitting in Figure 4(b) is difficult to distinguish from noise in Figure 5(a), while the noise in
Figure 5(b) completely overwhelms the Zeeman splitting.
Figures 6(a) and 6(b) show how Zeeman splitting in the largely-radial magnetic field at the discharge
channel exit can be reduced by rotating the laser polarization vector. Both spectra were recorded at the
8
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
(a) ṁA = 5 mg/s
(b) ṁA = 15 mg/s
Figure 5. Optogalvanic (black line) and cathode exit plane LIF (red line) spectra at 823.163 nm (air), thruster
operating at 150 V discharge voltage.
(a) π-polarization
(b) σ-polarization
Figure 6. Optogalvanic (black line) and discharge channel exit plane LIF (red line) spectra at 823.163 nm
(air), thruster operating at 150 V discharge voltage and ṁA = 5 mg/s.
9
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
discharge channel centerline, with the thruster operating at a 150 V discharge and an anode mass flow of
5 mg/s. The inner and outer coil currents were identical for both spectra. The laser polarization in Figure
6(a) was horizontal, so the polarization vector was parallel to the mostly-radial magnetic field (E k B, or
π-polarization) at the interrogation volume. For Figure 6(b), we used the half-wave plate and polarizing cube
beamsplitter to rotate the laser polarization so that it was vertical, placing the polarization perpendicular
to the magnetic field (E ⊥ B, or σ-polarization). The resulting shift is slight, but unmistakeable: the main
and secondary peaks shown in Figure 6(a) show definite Zeeman splitting in Figure 6(b).
B.
LIF signal strength vs. discharge voltage
500
Axial Velocity, m/s
450
400
350
300
250
200
150
100
50
0
0
50
100
150
200
250
300
Discharge Voltage, V
Figure 7. Axial velocity of neutral particles at the discharge chamber exit as a function of discharge voltage
for a constant 10 mg/s anode flow rate.
Figure 7 shows the axial bulk velocity of metastable neutrals at the discharge chamber exit as a function
of discharge voltage. The ±41.17
√ m/s error bars correspond to the 50-MHz resolution of the LIF scans.
The trendline is proportional to VD ; this is reasonable if we model the neutral flow as an electrothermal
thruster, where the exhaust velocity varies as the square root of the stagnation temperature. Since convective
losses would scale linearly with discharge velocity for constant discharge current, the stagnation temperature
could
also be expected to increase linearly with discharge voltage. (Since ion velocity is also proportional to
√
VD , this behavior might also indicate entrainment.)
Figure 8(a) shows the magnitude of the central peak versus the discharge voltage. The magnitude was
found by fitting a quadratic curve to the region in each scan representing the central peak. For a low-noise
scan, this region includes all data points above half of the approximate peak value. For high noise scan, this
region includes all points that rises above the noise amplitude of the region surrounding the peak. As such,
no peak magnitude can be found if no peak is readily identifiable. This is the reason why no 300 V data is
included in this figure.
All scans in this survey were done at an anode flow rate of 10 mg/s. In order to save time and cut
thermal drift, the relatively noise-free 50-V LIF scans were carried out using an integration time constant
τ = 1 s on the lock-in amplifier. The remaining scans were carried out using a τ = 3 s. All other hardware
configurations were kept constant throughout these scans. Note that while great care is taken to minimize
thermal drift between these scans, a large uncertainty bracket (±25%) is included to show that thermal
drifting can be a significant problem in these scans. A trendline has been added to shows that the peak
magnitude decreases rapidly as the discharge voltage increased.
In order to compare SNRs for scans taken at varying values of the integration time constant τ , we define
10
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
(a) Magnitude
(b) SNR
Figure 8. Magnitude and scaled signal-to-noise ratio of neutral LIF signal at the discharge chamber exit as a
function of discharge voltage for a constant 10 mg/s anode flow rate.
a scaled SNR
Rs =
σs
σn
r
τ
τ0
(25)
where σs is the standard deviation of the signala , σn is the standard deviation of the noiseb , and τ0 is a
scaling time constant (defined as τ0 = 1 s). The statistical argument behind Equation 25 assumes that the
noise is broadband, which is generally true for scans done with time constants that are on the same order of
magnitude.
Figure 8(b) shows the scaled SNR as a function of the discharge voltage. A linear curve-fit is included
to show the general trend. As the discharge voltage is increased, the signal rapidly gets drowned out by the
noise. Uncertainty bracket was not included due to the complexity of defining uncertainty for scaled SNR.
When taken in combination, Figures 8(a) and 8(b) strongly indicate that the number of neutrals in
the metastable lower state (presumably proportional to the total number of neutrals at the exit plane)
decreases dramatically with increasing discharge voltage. As a result, the LIF signal (proportional to the
metastable population) is effectively drowned out by noise (roughly independent of discharge voltage) before
the discharge voltage reaches 300 V.
C.
LIF signal strength vs. anode mass flow rate
Figure 9 shows the axial bulk velocity of metastable neutrals at the discharge chamber exit as a function of
anode mass flow rate. As before, the ±41.17 m/s error bars correspond to the 50-MHz resolution of the LIF
scans. The trendline is approximately constant; this is again reasonable if we model the neutral flow as an
electrothermal thruster. At constant discharge voltage and increasing mass flow rate, the convective heat
loss rate would increase linearly but the stagnation temperature (and thus the bulk velocity) would remain
approximately constant.
Figure 10(a) shows the magnitude of the central peak versus the anode mass flowrate at a constant
discharge voltage of 150 V. This plot was processed using the same techniques described for Figure8(a).
Data for the mass flow rate of 17.5 mg/s was not included in this figure because the peak was difficult to
locate. We used a 1-s time constant for the 10 mg/s scan, and 3-s time constants for the remaining scans.
Once again, we made every effort to minimize thermal drift during these scans.
Figure 10(b) shows how the scaled SNR varies with anode mass flow rate at a constant discharge voltage
of 150 V. As before the SNR calculation and scaling we carried out in the same manner as described for
Figure 8(b) .
a In the absence of a priori knowledge of the signal, we used a 10-point box smoothing of the noisy signal to approximate
the noise-free signal.
b Noise was defined as the difference between the smoothed signal and the noisy signal.
11
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
700
Axial Velocity, m/s
600
500
400
300
200
100
0
0
5
10
15
20
Mass Flow Rate, mg/s
Figure 9. Axial velocity of neutral particles at the discharge chamber exit as a function of anode flow rate for
a constant 150 V discharge
(a) Magnitude
(b) SNR
Figure 10. Magnitude and scaled signal-to-noise ratio of neutral LIF signal at the discharge chamber exit as
a function of mass flow rate for a constant 150 V discharge.
12
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007
When taken in combination, Figures 9, 10(a) and 10(b) show a somewhat counter-intuitive trend: as
the anode mass flow rate increases, the neutral velocity remains constant and the number of metastable
neutrals decreases. If the metastable population scales with the total neutral population, this implies that
the propellant utilization must increase (independant of neutral speed) with mass flow rate.
V.
Conclusions
We have carried out LIF scans of the neutral xenon 6s 2 [3/2]01 metastable at the discharge channel of
a 6-kW Hall thruster. The neutral velocity tends to rise with the square root of the discharge voltage at
constant mass flow rate, but remains essentially independant of mass flow rate at constant discharge voltage;
both behaviors indicate that the neutral flow in a Hall thruster follows the same principles found in an
electrothermal thruster. Rapidly-deteriorating LIF signal strengths and signal-to-noise ratios indicate that
the neutral density decreases monotonically with both discharge voltage and with mass flow rate.
VI.
Acknowledgements
This research was supported by the Air Force Office of Scientific Research grants F49620-02-1-0051 and
F49620-01-1-0061 (Dr. Mitat Birkan is the contract monitor for both). The authors are greatly appreciative
of this support.
We would also like to thank the graduate and undergraduate students at PEPL for their assistance in
experimental setup and operation, and the Department’s technical support staff for their aid with facility
repairs and modifications.
References
1 Reid, B.M., Gallimore, A.D., and R.R. Hofer, “Review of Hall thruster neutral flow dynamics,” 30th International
Electric Propulsion Conference, IEPC-2007-38, September 2007.
2 Linnell, J.A. and A.D. Gallimore, “Internal plasma potential measurements of a Hall thruster using plasma lens focusing,”
41st Joint Propulsion Conference, AIAA-2005-4402, July 2005.
3 Dymond, J.H., and E.B. Smith, The Virial Coefficients of Pure Gases and Mixtures, a Critical Compilation, Oxford
University Press, New York, 1980,
4 Jackson, D.A. and M.C. Coulombe. “Isotope shifts in the arc spectrum of xenon,” Proceedings of the Royal Society of
London A338(1974) 277-298.
5 Suzuki, M., Katoh, K., and N. Nishimiya. “Saturated absorption spectroscopy of Xe using a GaAs semiconductor laser,”
Spectrochimica Acta A58(2002) 25192531.
6 Jackson, D.A. and M.C. Coulombe. “Hyperfine structure in the arc spectrum of xenon,” Proceedings of the Royal Society
of London A327(1972) 137-145.
7 E. Browne. Table of Isotopes, C. M. Lederer and V. S. Shirley, ed. 7th ed., Wiley, 1978.
8 S. Svanberg. Atomic and Molecular Spectroscopy: Basic Aspects and Practical Applications. 3rd ed., Springer-Verlag,
2001.
9 H. Kopfermann. Nuclear Moments. Academic Press, 1958.
10 C. C. Candler. Atomic Spectra and the Vector Model. Van Nostrand, 1964.
13
The 30th International Electric Propulsion Conference, Florence, Italy
September 17-20, 2007